<?xml version="1.0" encoding="UTF-8"?><xml><records><record><source-app name="Biblio" version="7.x">Drupal-Biblio</source-app><ref-type>47</ref-type><contributors><authors><author><style face="normal" font="default" size="100%">Andrea Fedele</style></author></authors></contributors><titles><title><style face="normal" font="default" size="100%">Explain and Interpret Few-Shot Learning</style></title><secondary-title><style face="normal" font="default" size="100%">Joint Proceedings of the xAI-2023 Late-breaking Work, Demos and Doctoral Consortium co-located with the 1st World Conference on eXplainable Artificial Intelligence (xAI-2023), Lisbon, Portugal, July 26-28, 2023</style></secondary-title></titles><dates><year><style  face="normal" font="default" size="100%">2023</style></year></dates><urls><web-urls><url><style face="normal" font="default" size="100%">https://ceur-ws.org/Vol-3554/paper38.pdf</style></url></web-urls></urls><publisher><style face="normal" font="default" size="100%">CEUR-WS.org</style></publisher><language><style face="normal" font="default" size="100%">eng</style></language><abstract><style face="normal" font="default" size="100%">Recent advancements in Artificial Intelligence have been fueled by vast datasets, powerful computing resources, and sophisticated algorithms. However, traditional Machine Learning models face limitations in handling scarce data. Few-Shot Learning (FSL) offers a promising solution by training models on a small number of examples per class. This manuscript introduces FXI-FSL, a framework for eXplainability and Interpretability in FSL, which aims to develop post-hoc explainability algorithms and interpretableby-design alternatives. A noteworthy contribution is the SIamese Network EXplainer (SINEX), a post-hoc approach shedding light on Siamese Network behavior. The proposed framework seeks to unveil the rationale behind FSL models, instilling trust in their real-world applications. Moreover, it emerges as a safeguard for developers, facilitating models fine-tuning prior to deployment, and as a guide for end users navigating the decisions of these models</style></abstract></record><record><source-app name="Biblio" version="7.x">Drupal-Biblio</source-app><ref-type>47</ref-type><contributors><authors><author><style face="normal" font="default" size="100%">Andrea Fedele</style></author><author><style face="normal" font="default" size="100%">Riccardo Guidotti</style></author><author><style face="normal" font="default" size="100%">Dino Pedreschi</style></author></authors></contributors><titles><title><style face="normal" font="default" size="100%">Explaining Siamese Networks in Few-Shot Learning for Audio Data</style></title><secondary-title><style face="normal" font="default" size="100%">Discovery Science - 25th International Conference, DS 2022, Montpellier, France, October 10-12, 2022, Proceedings</style></secondary-title></titles><dates><year><style  face="normal" font="default" size="100%">2022</style></year></dates><urls><web-urls><url><style face="normal" font="default" size="100%">https://doi.org/10.1007/978-3-031-18840-4_36</style></url></web-urls></urls><publisher><style face="normal" font="default" size="100%">Springer</style></publisher><language><style face="normal" font="default" size="100%">eng</style></language><abstract><style face="normal" font="default" size="100%">Machine learning models are not able to generalize correctly when queried on samples belonging to class distributions that were never seen during training. This is a critical issue, since real world applications might need to quickly adapt without the necessity of re-training. To overcome these limitations, few-shot learning frameworks have been proposed and their applicability has been studied widely for computer vision tasks. Siamese Networks learn pairs similarity in form of a metric that can be easily extended on new unseen classes. Unfortunately, the downside of such systems is the lack of explainability. We propose a method to explain the outcomes of Siamese Networks in the context of few-shot learning for audio data. This objective is pursued through a local perturbation-based approach that evaluates segments-weighted-average contributions to the final outcome considering the interplay between different areas of the audio spectrogram. Qualitative and quantitative results demonstrate that our method is able to show common intra-class characteristics and erroneous reliance on silent sections.</style></abstract></record></records></xml>